paper-extractor.py

import os

from pdfminer.pdfinterp import PDFResourceManager, PDFPageInterpreter
from pdfminer.converter import TextConverter
from pdfminer.layout import LAParams, LTRect, LTLine, LTTextBoxHorizontal
from pdfminer.pdfpage import PDFPage
from cStringIO import StringIO
from pdfminer.converter import PDFPageAggregator

from wand.image import Image

laparams = LAParams()
rsrcmgr = PDFResourceManager()
document = file('c3u.pdf', 'rb')
device = PDFPageAggregator(rsrcmgr, laparams=laparams)
interpreter = PDFPageInterpreter(rsrcmgr, device)

for i, page in enumerate(PDFPage.get_pages(document)):
    # if i == 3:
    #     break

    print "Page {}".format(i)

    interpreter.process_page(page)
    layout = device.get_result()
    # print layout._objs
    
    space = 0
    textboxes = [r for r in layout._objs if type(r) is LTTextBoxHorizontal ]
    for t in textboxes:
        if t.get_text().startswith("Answer space for question"):
            space = t.y0

    path = os.path.join(os.getcwd(), "c3u.pdf[{}]".format(i))

    scale = 3

    with Image(filename=path, resolution=int(72*scale)) as img:
        # Crop the margins
        x = 46 * scale
        y = 66 * scale
        width = 489 * scale
        height = 711 * scale
        height -= max((space - 51) * scale, 1)
        img.crop(int(x), int(y), width=int(width), height=int(height))
        img.save(filename="test/test{}.jpg".format(i))

• Remove front page
• Remove blank pages
• Handle drawing boxes
• Test more papers